Syntheses were carried out in an oxygen and moisture-free atmosphere. Solvents were dried by reflux over the appropriate drying agent and distilled under a stream of argon. Fc, Fc-NH₂, 1-hydroxy-benzotriazole hydrate (HOBt), diisopropylethylamine (DIEA), succinic anhydride, glutaric anhydride and pimelic acid were Sigma Aldrich products; 1-(3-dimethylaminopropyl)-3-ethylcarbodiimide (EDC), and L-alanine ethyl ester hydrochloride (H-L-Ala-OEt·HCl) were Iris Biotech products. Boc-Gly-L-Ala-OEt was prepared by literature procedures.

MS spectra were obtained using an Agilent 6130 ESI-TOF mass spectrometer coupled to an HPLC system, collecting data in the positive mode. All compounds were analyzed by analytical HPLC and were >95% pure.

Spectra were collected on a Bruker Avance III HD spectrometer operating at 400.13 MHz (T = 298K). The peptide concentration in solution was 1 mM in spectrograde CHCl₃-d₁ (99.8% D containing 0.5 %wt. of silver foils as stabilizers and 0.03% (v/v) tetramethylsilane–Eurisotop) and DMSO-d₆ (99.96% D–Eurisotop). Processing and evaluation of the experimental data were carried out using the TOPSPIN software packages. All homonuclear spectra were acquired by collecting 400 experiments, each consisting of 32 scans and 2K data points. The spin systems of the amino acid residues were identified using standard double-quantum filtered COSY (Rance et al., 1983) and clean TOCSY (Griesinger et al., 1988) spectra. In the latter case, the spin-lock sequence was 70 ms long. NOESY experiments were utilized for sequence-specific assignments (Wüthrich, 1991; Boros et al., 2018).

Data for 1 and N-ferrocenylsuccinimide (Fc-Su), obtained as byproduct in the synthesis of 2, were collected using an Oxford Diffraction Gemini E diffractometer, equipped with a 2 K × 2 K EOS CCD area detector and sealed–tube Enhance (Cu-Kα, λ = 1.54178 Å) X–ray source. Data for compound 3 were collected on a Bruker D8 Venture diffractometer equipped with Incoatec IμS3.0 (EF) microfocus sealed-tube (Cu-Kα, λ = 1.54178 Å), a Montel layer optics monochromator, and a Photon III C14 CPAD area detector. Data reduction, finalization and cell refinement were carried out through the CrysAlisPro software. Empirical multi-scan absorption corrections using equivalent reflections have been performed with the scaling algorithm SCALE3 ABSPACK. Accurate unit cell parameters were obtained by least squares refinement of the angular settings of strongest reflections, chosen from the whole experiment. The structures were solved with Olex2 (Dolomanov et al., 2009) by using ShelXT structure solution program by Intrinsic Phasing and refined with the ShelXL refinement package using least-squares minimization (Sheldrick, 2015). In the last cycles of refinement, non-hydrogen atoms were refined anisotropically. Hydrogen atoms were included in calculated positions, and a riding model was used for their refinement. Supplementary Table S1 reports the crystal data and refinement details. Cambridge Crystallographic Data Centre (CCDC) numbers 2497229-2497,231 contain the supplementary crystallographic data for this paper. These data are provided free of charge by the joint CCDC and Fachinformationszentrum Karlsruhe Access Structures service www.ccdc.ca-m.ac.uk/structures.

The experiments were performed in an air-tight three-electrode cell connected to a vacuum/argon Schlenk line. Dichloromethane solvent was pre-dried with anhydrous calcium chloride, refluxed over calcium hydride and distilled under a stream of Argon. Solvent and nBu₄NPF₆ were degassed in a Schlenk flask by manifold freeze-pump-thaw cycles and transferred by cannula in the cell. The reference electrode was a SCE (Tacussel ECS C10) separated from the solution by a bridge compartment filled with the same solvent/supporting electrolyte solution used in the cell. The counter electrode was a platinum spiral with around 1 cm² apparent surface area. The working electrode was a disk obtained from cross section of a gold wire with 0.5 and 0.125 mm diameter sealed in glass. Between successive scans, the working electrode was polished on alumina according to standard procedures and sonicated before use. An, E.G.,&G PAR-175 signal generator was used. The currents and potentials were recorded on a Lecroy 9310L oscilloscope. The potentiostat was home-built with a positive feedback loop for compensation of the ohmic drop (Amatore et al., 1989).

The peptides were dissolved in DMSO at a concentration of 5 mg/mL. In Eppendorf tubes, 1 mL of HEPES buffer (25 mM, pH = 7.6) was temperature equilibrated at 37 °C before adding 250 μL of human serum and 20 μL of peptide solution. The human serum used in these experiments was obtained from Merck. It is derived from platelet-poor human plasma. It is sterile-filtered and tested for viruses and mycoplasma. The reaction was monitored for 24 h. At fixed intervals, 100 μL of the solution was withdrawn and added with 200 μL of absolute ethanol. The sample was kept on ice for 15 min and then centrifuged at 16,000 g for 5 min. Finally, the supernatant solution was analyzed by HPLC or HPLC–MS. (Jenssen and Aspmo, 2008). A peptide of similar length, not resistant to degradation in serum, was used as a positive test. The peptide sequence is as follows: Tyr-Ser-Ser-Phe-Leu. To verify peptide stability in buffer solution, samples containing peptide solutions, buffer, and ethanol were also analyzed.

Compounds 1-6 (Table 1) were synthesized by solution methods and characterized by HPLC-MS and NMR spectrometry as described in detail in the Supplementary Material file. Briefly, peptide bonds were formed by activating the carboxylic function with the coupling reagents EDC/HOBt [EDC, N-ethyl-N′-dimethylaminopropyl-carbodiimide; HOBt, 1-hydroxybenzotriazole]. HOBt suppresses racemization, when chiral amino acids are involved, but it also catalyzes the formation of amide bonds (Crisma et al., 1998). For this latter reason it was used even when chirality was not an issue. The amide bonds of compounds 1 and 3 were prepared by reacting Fc-NH₂ with succinic and glutaric anhydride, respectively. In the case of compound 5, Fc-NH₂ was first reacted with pimelic acid. The product obtained was isolated and, without further purification, reacted with H-Gly-Gly-Ala-OEt (see Supplementary Material).

When 1 was activated with EDC/HOBt to synthesize 2, we also obtained N-ferrocenylsuccinimide (Fc-Su) as a byproduct, due to an intramolecular amide (imide) bond formation (Supplementary Material). The same reaction occurred in the synthesis of 4, starting from 3, yielding a moderate amount of N-ferrocenylglutarimide (Supplementary Material). We were able to grow single crystals of both Fc-Su and its precursor, compound 1, and to determine their molecular structures by X-ray diffraction (Figure 3).

The structure of 1 was solved in the monoclinic space group P2₁/c, while that of Fc-Su was solved in the monoclinic space group P2₁/m. In both structures, the cyclopentadienyl rings coordinate to the Fe(II) ion with η⁵ hapticity, adopting a nearly eclipsed configuration. Such an arrangement has been previously observed in several ferrocene derivatives (Donoli et al., 2013; Santi et al., 2021; Biondi et al., 2021; Biondi et al., 2022). The Fe–C bond lengths range from 2.025 to 2.045 Å, while the distances between the Fe(II) ion and the centroids of the cyclopentadienyl rings are between 1.642-1.644 Å.

Interestingly, in 1 the angle between the plane of the cyclopentadienyl ring (C6–C10) and the plane defined by atoms N1–C11–O1–C12 is 8.3°, whereas in Fc-Su, the angle between the cyclopentadienyl plane (C4–C5–C6–C5′–C4’) and the succinimide plane is only 0.2°. This coplanarity suggests that an extended π-conjugation is established between Cp and the imide of Fc-Su. Such conjugation exists also between Cp and the amide in 1, although to a lower extent.

In compound 1, the N1–H1 and O3–H3 groups participate in intermolecular hydrogen bonds with atoms O2 and O1, respectively, as described in Table 2 and shown in Figure 4.

We were also able to solve the crystal structure of Boc-Gly-Gly-Ala-OEt, the tripeptide intermediate used in the synthesis of compound 6. The structure was determined in the monoclinic space group P2₁. The asymmetric unit comprises two crystallographic independent molecules, designated A and B (Figure 5A). Both molecules feature a i → i+3 intramolecular C=O⋅⋅⋅H–N hydrogen bond, which stabilizes a β-turn conformation. In molecule A, the φ and ψ dihedral angles adopt values close to those characteristics of a type III’ β-turn, whereas molecule B exhibits negative angles consistent with a type III β-turn (Table 3) (Venkatachalam, 1968). The intramolecular hydrogen bond forms a 10-membered pseudo-cycle with very similar geometries in both molecules, with N···O donor–acceptor distances of approximately 3.0 Å for A (N3A⋅⋅⋅O2A) and B (N3B⋅⋅⋅O2B) (Table 4).

Intermolecular H-bonds connect molecules A and B into a 1D supramolecular chain, arranged in an alternating A–B sequence (Figure 5C). This motif is sustained by N1A⋅⋅⋅O3B and N2A⋅⋅⋅O4B H-bonds, along with the reciprocal N1B⋅⋅⋅O3A and N2B⋅⋅⋅O4A interactions (Table 4).

This β-turn structure was rather unexpected, as Gly usually prefers to adopt extended conformations. We may argue that the β-turn was promoted by the crystal packing and/or by the achiral nature of Gly. Indeed, the two independent molecules exhibit opposite helical handedness and form a pseudo-racemate: molecule A twists anticlockwise and molecule B clockwise (Figure 5B). It is well known that racemates and pseudo-racemates crystallize easily and facilitate structure resolution (Zawadke and Berg, 1993; Doi et al., 1993; Toniolo et al., 1994). However, we are inclined to believe that the observed bended conformation of Boc-Gly-Gly-Ala-OEt does not depend on packing or chirality because it is maintained in solution and also when the tripeptide is linked to Fc. As a matter of fact, typical cross-peaks of turn/helical peptide conformations (Wüthrich, 1991) are observed in the NOESY spectrum of 6 in CDCl₃ solution. Both NH _i → NH _i+1 connectivities characterizing the β-turn are clearly detected (Figure 6). Since the β-turn involves the two achiral Gly residues, the left- and right-turn form likewise with the same probability. However, we are unable to assess if one conformation prevails over the other because they interconvert at a much faster rate than the NMR time scale (Hummel et al., 1987).

Moreover, the stable presence in solution of such β-turn could account for the fair resistance to enzymatic degradation of compound 6 (see below).

Fc is a very useful redox probe as it is stable in water, it is insensitive to air in both neutral and oxidized states, it is soluble in common organic solvents and in addition it displays a low oxidation potential. As previously pointed out, the way Fc binds to an organic moiety significantly influences its redox potential. To better understand this issue, we prepared the above-mentioned Fc-conjugates, that differ for spacer and peptide length, but share the way Fc is connected, i.e., the amide nitrogen (Fc-NH-CO-). The most exploited way to conjugate Fc to a peptide for creating new drugs has been so far Fc-CO-NH-. With our new molecules, we expect to lower the Fc redox potentials below the original ferrocene values, and thus make them compatible with the intracellular redox potentials, which range from +0.40 V to −0.44 V vs. SCE.

The cyclic voltammograms of the peptides N-Fc-succinamide-Gly-Ala-OEt (2), N-Fc-glutaramide-Gly-Ala-OEt (4), N-Fc-pimelamide-Gly-Ala-OEt (5), N-Fc-glutaramide-Gly-Gly-Ala-OEt (6), and of the carboxylic acids Fc (1) and Fc-NH ₂ (3) are reported in Figure 7. All measurements were performed in dichloromethane solution in the presence of (nBu)₄BPF₆ (0.1 M) as a supporting electrolyte.

The main objective was to evaluate the effect exerted on the oxidation potential (E _1/2) by the -NHCO- group, the spacer and the peptide and consequently to assess the suitability of these compounds to meet the intracellular redox conditions. The electron-donating properties of the -NHCO- group determine a decrease of the oxidation potential of Fc down to 0.3 V for some of our molecules (Table 5). On the contrary, the electron-withdrawing effect of the -CONH- group causes an increase in the oxidation potential up to values of 0.6–0.8 V (see, for example, Biondi et al., 2022), which are not compatible with the intracellular environment (between 0.40 and −0.44 V vs. SCE).

It can be observed that all the products analyzed show a single, reversible, Nernstian monoelectronic oxidation wave (ΔEp = 60–70 mV, Table 5) in the scan rate v range between 0.1 and 5 Vs ^-1, on electrode gold disc working with diameter d = 0.125 mm (0.5–5 Vs ^-1) and 0.5 mm (0.1–0.5 Vs^-1).

As expected, the E _1/2 values demonstrate that the Fc-NH-CO- linkage, i.e., Fc directly connected to N, has the major effect on the oxidation potential of our Fc-peptides, anticipating by 100–120 mV with respect to Fc, with the greatest decrease obtained for peptide 5. Apparently, a long spacer helps decrease the E _1/2 values, probably because it moves far apart the second, electron-withdrawing carbonyl. This conclusion agrees with the behavior of 1 and 3, as the latter endowed with a longer chain displays a 30 mV reduction of E _1/2.

The length of the peptides also appears to have an effect, albeit a moderate one. The E _1/2 value of 6, containing a tripeptide, is slightly lower than in the case of 4, a conjugate with a dipeptide.

Overall, all Fc-peptides (2 and 4–6) have shown to possess redox values compatible with the intracellular potentials.

The stability of Fc-peptide conjugates against proteolytic degradation is a major challenge in their development as potential anticancer agents. In this study, we investigated the stability of the synthetic peptides when exposed to human serum. After incubation with serum, the integrity of the conjugates was monitored by HPLC. Proteolytic degradation is reported as a function of time (Peggion et al., 2025; Albini, et al., 2025).

Considering the dipeptide Gly-Ala conjugated to Fc via linkers of different lengths, we observed rapid degradation of 4, whereas 2 exhibited a slower degradation rate (Figure 8). We concluded that the Fc moiety exerts a protective effect on the Gly-Ala sequence, depending on the distance between the Fc moiety and the peptide.

Surprisingly, a notably different pattern of degradation was observed for 6. Here, the peptide moiety is represented by the Gly-Gly-Ala sequence, and the Fc unit is connected to the peptide via the same linker as in compound 4. In this case, the percentage of intact peptide was around 60% after 3 h, and almost 30% of the peptide was still present after 24 h (data not shown). The conformational analysis conducted on the free peptide and on 6 revealed the presence of a β-turn structure. This structural feature led us to interpret the observed resistance to proteolysis as an effect of conformational stability rather than the protective role of the Fc moiety.